Cancer Metabolism

1. The Warburg Effect — Aerobic Glycolysis

In 1924, working with thinly sliced rat hepatoma tissue in saline-buffered Ringer’s solution, Otto Warburg made a measurement that has shaped cancer biology for a century. Tumour slices, even when generously supplied with oxygen, fermented glucose to lactate at rates an order of magnitude higher than the matched normal tissue (Warburg, Klin Wochenschr 1924; consolidated in Science 1956).

In normal differentiated cells the disposal of pyruvate is dictated by oxygen: under aerobic conditions pyruvate enters the mitochondrion and is oxidised to CO₂ via the TCA cycle and oxidative phosphorylation; only under hypoxia does it overflow to lactate (the Pasteur effect). Tumour cells short-circuit this rule. Even in fully oxygenated tissue, a large fraction of glucose-derived pyruvate is reduced to lactate by lactate dehydrogenase A (LDH-A) and excreted — hence aerobic glycolysis.

\[ \text{glucose} + 2\,\text{ADP} + 2\,P_i \;\xrightarrow{\text{glycolysis}}\; 2\,\text{lactate} + 2\,\text{ATP} + 2\,\text{H}_2\text{O} \]

Compared to complete oxidation of glucose \( (\Delta G^{\circ\prime} \approx -2870 \,\text{kJ/mol}, \;\sim 30\text{–}32\,\text{ATP}) \) fermentative lactate production yields only \( 2\,\text{ATP/glucose} \). The puzzle Warburg posed was therefore thermodynamic and teleological: why does a fast-growing cell choose the less-efficient pathway? Warburg himself believed the answer was a defectin mitochondrial respiration — an idea that proved largely incorrect, since most cancer cells have functional mitochondria.

The modern reframing was articulated by Vander Heiden, Cantley & Thompson (Science 2009): aerobic glycolysis is not about ATP yield but about the flux of carbon and reducing power into biosynthesis. A proliferating cell must double its dry mass before it can divide, which requires nucleotides, amino acids, and lipids — and these come from glucose-derived intermediates, not from CO₂.

At the molecular level the Warburg phenotype is enforced by a coordinated programme: up-regulation of glucose transporters (GLUT1/SLC2A1), hexokinase II (HK2), phosphofructokinase (PFK1, PFKFB3), the pyruvate-kinase M2 isoform (PKM2), LDH-A and the monocarboxylate exporter MCT4 (SLC16A3). Master regulators are HIF-1α (oxygen- and oncogene-driven) and MYC.

2. Biosynthetic Rationale — Why Aerobic Glycolysis Makes Sense

Once one accepts that the relevant currency for a proliferating cell is not ATP but carbon and reducing equivalents, the Warburg solution becomes coherent. Glycolytic flux has three biosynthetic destinations:

The PKM2 paradox

The penultimate glycolytic enzyme pyruvate kinase exists as four isoforms (L, R, M1, M2). Embryonic tissues and tumours preferentially express the M2 isoform (PKM2), which — unlike the constitutively active M1 — is allosterically gated by fructose-1,6-bisP and tyrosine-phosphorylation signals downstream of growth-factor receptors. Christofket al. (Nature 2008) showed that PKM2 can be held in a low-activity state, causing upstream glycolytic intermediates to back up and spill into the PPP, serine and hexosamine branches.

\[ \text{biomass yield per glucose} \;\propto\; \frac{[\text{glycolytic flux}]}{[\text{PKM2 activity}]} \]

Counter-intuitively, increasing PKM2 activity (with small-molecule activators such as TEPP-46) slows proliferation by removing the biosynthetic spillover. The Warburg effect is thus an engineered bottleneck.

3. Glutaminolysis and the MYC Connection

Glucose alone cannot fuel proliferation. A cancer cell must also import nitrogen and carbon for amino-acid and nucleotide biosynthesis — and for many tumour types the principal source is glutamine, the most abundant amino acid in plasma (~0.5 mM). DeBerardinis and colleagues (Cell Metab 2008; PNAS 2007) showed that MYC-transformed cells are “glutamine-addicted”: glutamine withdrawal triggers rapid apoptosis even with abundant glucose.

Glutaminolytic flux

• SLC1A5 (ASCT2) imports glutamine; SLC7A11 exchanges it for cystine in the xCT antiporter.
• Glutaminase (GLS, kidney-type GLS1) deamidates glutamine to glutamate in the mitochondrial matrix, releasing NH₄⁺.
• Glutamate is then converted to α-ketoglutarate (α-KG) by glutamate dehydrogenase (GLUD1) or transaminases (GOT1, GPT2), feeding the TCA cycle.
• α-KG can flow oxidatively (forward TCA → succinate → OAA) or reductively (IDH1/2 → isocitrate → citrate → cytosolic acetyl-CoA) under hypoxia or mitochondrial dysfunction.

MYC as glutaminolytic master regulator

MYC (and MYCN in neuroblastoma) directly transactivates SLC1A5, GLS1 (by repressing its negative regulator miR-23a/b), GLUD1, and the de-novo nucleotide biosynthetic enzymes that consume glutamine-derived nitrogen. The result is a cell wired to run both glycolysis and glutaminolysis in parallel: glucose feeds anabolic carbon, lactate and lipogenesis; glutamine feeds anaplerotic TCA flux, NADPH (via malic enzyme) and non-essential amino-acid synthesis.

\[ \text{glutamine} \xrightarrow{\text{GLS}} \text{glutamate} \xrightarrow{\text{GLUD/GOT}} \alpha\text{-KG} \rightarrow \text{TCA} \rightarrow \text{OAA, aspartate, pyrimidines} \]

4. The Lipogenic Switch

Cancer cells must synthesise membrane lipids de novo. Most adult tissues import fatty acids from circulation; tumour cells, by contrast, re-activate the foetal/hepatic lipogenic programme. The defining metabolic re-routing is cytosolic citrate → acetyl-CoA → palmitate:

• ATP-citrate lyase (ACLY) cleaves cytosolic citrate (exported from mitochondria via SLC25A1) to acetyl-CoA + OAA. ACLY is acetylated and stabilised in many tumours and is a key node sensing glucose/glutamine flux.
• Acetyl-CoA carboxylase (ACC1/ACC2) carboxylates acetyl-CoA to malonyl-CoA — the committed, rate-limiting step. ACC is inhibited by AMPK phosphorylation (S79/S221) and by long-chain fatty acyl-CoA feedback.
• Fatty-acid synthase (FASN) — a ~270 kDa homodimeric multi-enzyme — condenses 7 malonyl-CoA + 1 acetyl-CoA, using 14 NADPH, into one C16:0 palmitate (see PDB 5LF1 below).
• SCD1 introduces the Δ9 double bond (palmitate → palmitoleate; stearate → oleate), with ELOVL elongases producing C18–C24 species for ER membrane fluidity, sphingolipids, and signalling.

\[ \text{acetyl-CoA} + 7\,\text{malonyl-CoA} + 14\,\text{NADPH} \xrightarrow{\text{FASN}} \text{palmitate (C16:0)} + 7\,\text{CO}_2 + 8\,\text{CoA} + 14\,\text{NADP}^+ + 6\,\text{H}_2\text{O} \]

SREBP-1c as transcriptional master switch

The lipogenic genes (ACC, FASN, SCD1, ACLY) are coordinately controlled by SREBP-1c (sterol-regulatory-element binding protein 1c), processed from its ER precursor by S1P/S2P proteases when sterols are low (or in response to PI3K–AKT–mTORC1 signalling). In cancer, mTORC1 drives SREBP-1c activation downstream of constitutive PI3K signalling, while SREBP-2 handles cholesterol biosynthesis (HMGCR, SQLE).

FASN over-expression is detectable in the majority of breast, prostate, colorectal, ovarian and hepatocellular carcinomas, where it correlates with poor prognosis. The first-in-class FASN inhibitor TVB-2640 (denifanstat) has shown activity in NASH and is in oncology trials.

5. Oncometabolites — Metabolites that Drive Cancer

The most striking conceptual advance of the modern metabolism era is the recognition that metabolites themselves can be oncogenic. In three connected discoveries, mutations in TCA-cycle enzymes were shown to cause pathological accumulation of small molecules that act as oncometabolites by competitive inhibition of α-ketoglutarate-dependent dioxygenases.

5.1 IDH1/2 and (R)-2-hydroxyglutarate

Wild-type isocitrate dehydrogenase (IDH1 cytosolic, IDH2 mitochondrial) catalyses isocitrate → α-KG + CO₂ + NADPH. Yan et al. (NEJM 2009) found heterozygous IDH1 R132H and IDH2 R140Q/R172K mutations in ~80% of grade II/III gliomas and secondary glioblastomas, and a smaller fraction of AML and chondrosarcoma. The mutant enzyme acquires a neomorphic activity: instead of producing α-KG it reduces it further to (R)-2-hydroxyglutarate (R-2HG), consuming NADPH (Dang et al., Nature 2009).

\[ \text{isocitrate} \xrightarrow{\text{IDH wt}} \alpha\text{-KG} \xrightarrow{\text{IDH R132H}} (R)\text{-2HG} \]

R-2HG accumulates to millimolar intracellular concentrations — up to 10 mM in IDH-mutant gliomas, vs <100 µM in normal tissue. Because R-2HG is structurally near-isosteric with α-KG, it competitively inhibits the entire family of α-KG-dependent Fe(II) dioxygenases, with K_i in the µM–mM range matching the achieved tumour concentration.

5.2 Succinate (SDH loss) and fumarate (FH loss)

Two TCA-cycle tumour-suppressor genes follow the same logic. SDHA/B/C/D (succinate dehydrogenase, Complex II) loss causes succinate accumulation; biallelic loss drives hereditary paraganglioma/phaeochromocytoma and a distinct subset of GIST. FH (fumarate hydratase) loss causes fumarate accumulation in hereditary leiomyomatosis and renal-cell carcinoma (HLRCC, Tomlinson et al. 2002). Like 2HG, both succinate and fumarate competitively inhibit α-KG-dependent dioxygenases.

5.3 The α-KG-dependent dioxygenase family — the common target

These ~70 human enzymes share a Fe(II)/2-oxoglutarate active-site geometry. They split molecular O₂, decarboxylating α-KG to succinate and CO₂ while oxidising the substrate. The cancer-relevant subfamilies are:

The dual mechanism — epigenetic via TET and JmjC inhibition, and pseudohypoxic via PHD inhibition — explains nearly every phenotype of IDH-, SDH- and FH-mutant tumours:

• CIMP (CpG-island methylator phenotype): IDH-mutant gliomas, AMLs and chondrosarcomas show genome-wide hypermethylation due to TET2 inhibition (Figueroa et al. Cancer Cell 2010; Turcanet al. Nature 2012).
• Differentiation block: in AML, mutant IDH and TET2 are mutually exclusive driver lesions — both block hydroxymethylation-dependent enhancer activation needed for myeloid differentiation.
• Pseudohypoxia: in SDH/FH-mutant tumours, PHD inhibition stabilises HIF-1α/2α even in oxygenated tissue, switching on the canonical hypoxic transcriptome (GLUT1, VEGF, EPO, glycolytic enzymes).
• Aberrant histone marks: increased H3K9 and H3K27 trimethylation due to KDM4/KDM6 inhibition; impairs lineage-specifying enhancers.
• HR-repair deficiency: 2HG and fumarate also inhibit lysine demethylases at H3K9me3, suppressing HR and conferring BRCAness in IDH-mutant tumours — rationale for PARP-i sensitivity (see Part IV).

6. The TCA Cycle in Cancer — Anaplerosis & Reductive Carboxylation

In a quiescent cell the TCA cycle (background) is a closed catalytic loop oxidising acetyl-CoA to CO₂. In a proliferating cell, intermediates are continuously withdrawn for biosynthesis — oxaloacetate for aspartate and pyrimidines, α-KG for non-essential amino acids, succinyl-CoA for haem, citrate for cytosolic acetyl-CoA. The cycle therefore needs continuous anaplerotic refilling.

Anaplerotic inputs

• Glutamine → α-KG via glutaminase + transaminase. The dominant input in MYC- and KRAS-driven tumours.
• Pyruvate → OAA via pyruvate carboxylase (PC). Up-regulated in NSCLC and some breast cancers as a glutamine-independent route.
• Branched-chain amino acids (Leu, Ile, Val) → acetyl-CoA / succinyl-CoA via BCAT and BCKDH; BCAT1 is up-regulated in IDH-wild-type AML.
• Acetate → acetyl-CoA via ACSS2; an emerging fuel in brain metastases and lipogenic tumours under stress.

Reductive carboxylation

Under hypoxia or with mitochondrial dysfunction (e.g. fumarase loss, complex-III inhibition), cells run a portion of the TCA cycle in reverse:

\[ \alpha\text{-KG} + \text{CO}_2 + \text{NADPH} \xrightarrow{\text{IDH1/IDH2 (reverse)}} \text{isocitrate} \xrightarrow{\text{aconitase}} \text{citrate} \xrightarrow{\text{ACLY}} \text{acetyl-CoA (cytosolic)} \]

This reductive carboxylation (Metallo et al. Nature 2011; Mullen et al. Nature 2011) allows hypoxic cancer cells to derive cytosolic acetyl-CoA from glutamine instead of glucose — sustaining lipogenesis when glycolytic carbon cannot reach the mitochondrion. Stable-isotope tracing with [5-¹³C]-glutamine reveals the M+5 citrate/M+3 acetyl-CoA labelling pattern that diagnoses this flux.

Importantly, the same enzymes (IDH1, IDH2) that drive the oncometabolite phenotype in their mutant form catalyse reductive carboxylation in their wild-type form — another illustration of why a non-redundant metabolic node, once mutated, can have such pleiotropic phenotypic consequences.

7. mTORC1 and Anabolic Reprogramming

The biosynthetic programme described above is integrated by mTORC1 (mechanistic target of rapamycin complex 1), the central anabolic kinase of the eukaryotic cell. mTORC1 senses growth factors (PI3K–AKT → TSC1/TSC2 → Rheb-GTP), amino acids (Rag GTPases loaded by GATOR2/CASTOR/Sestrin), energy status (AMPK → TSC2 / Raptor S722, S792), and oxygen (REDD1).

mTORC1 outputs — the anabolic programme

• Protein synthesis — phosphorylation of S6K1 (T389) and 4E-BP1 (T37/46, T70, S65) drives cap-dependent translation; ribosome biogenesis via TIF-IA and POL-I.
• Lipid synthesis — processing of SREBP-1c, induction of FASN/ACC/SCD1.
• Nucleotide synthesis — S6K1 phosphorylates CAD (T1406) for pyrimidine biosynthesis; mitochondrial 1C-flux for purines.
• Glycolysis — HIF-1α stabilisation under normoxia; PKM2 and HK2 induction.
• Autophagy suppression — ULK1 S757 phosphorylation blocks autophagosome initiation; transcription factor TFEB cytoplasmic sequestration.

Genetic activation of mTORC1 in cancer occurs through PI3K/AKT pathway lesions (PIK3CA, PTEN, AKT1) and through TSC1/TSC2 loss in tuberous sclerosis-associated tumours. The PI3K–AKT–mTORC1 axis is the most frequently mutated pathway across all human cancers.

\[ \text{[ATP]/[ADP]} \downarrow \;\Rightarrow\; \text{AMPK active} \;\Rightarrow\; \text{TSC2}^{S1387\text{-P}}, \text{Raptor}^{S792\text{-P}} \;\Rightarrow\; \text{mTORC1 off} \;\Rightarrow\; \text{anabolism off, autophagy on} \]

AMPK — the cellular energy gauge, activated allosterically by AMP and through phosphorylation by LKB1 (T172) — acts as the antagonist of mTORC1. The tumour-suppressor LKB1/STK11 (lost in Peutz-Jeghers and 30% of NSCLC) sits upstream; metformin partially activates AMPK indirectly via mild Complex-I inhibition.

8. Tumour pH, Lactate Shuttle, and the Reverse Warburg Effect

A glycolytic tumour cell exports two protons per glucose alongside lactate, plus additional H⁺ from CO₂/HCO₃⁻ via membrane carbonic anhydrases (CA9, CA12). The result is a paradoxical pH gradient: extracellular tumour pH falls to 6.5–7.0, while intracellular pH is maintained alkaline at 7.2–7.4 — the reverse of normal tissue.

\[ \Delta\text{pH}_\text{tumour} \;=\; \text{pH}_\text{i} - \text{pH}_\text{e} \;\approx\; +0.4 \quad \text{(normal tissue: } \approx -0.1\text{)} \]

Acid-extruding machinery

• MCT1 (SLC16A1) — bidirectional H⁺/lactate co-transporter; favoured for lactate import.
• MCT4 (SLC16A3) — HIF-1-induced; high V_max, lactate exporter.
• NHE1 (SLC9A1) — Na⁺/H⁺ exchanger; first responder to acidification.
• V-ATPase — sequesters protons in lysosomes/exosomes; surface V-ATPase reported in metastatic cells.
• Carbonic anhydrase IX/XII (CA9/CA12) — HIF-induced surface enzymes that hydrate extracellular CO₂ to HCO₃⁻ + H⁺, buffering intracellular space.

The lactate shuttle

Lactate is no longer regarded as a metabolic waste. Following Sonveaux et al. (JCI 2008) and Hui et al. (Nature 2017), tumour lactate is recognised as a primary respiratory fuel: oxygenated tumour zones import lactate via MCT1, oxidise it to pyruvate (LDH-B), and feed it into the TCA cycle — sparing glucose for hypoxic glycolytic cells deeper in the tumour. This metabolic symbiosis mirrors the astrocyte–neuron lactate shuttle of brain tissue.

The “reverse Warburg effect”

Pavlides et al. (Cell Cycle 2009) proposed that, in some breast and prostate carcinomas, the stromal cancer-associated fibroblasts run aerobic glycolysis (loss of stromal Cav-1, oxidative stress) and supply lactate and pyruvate to the oxidative tumour cells. The cancer cell becomes the consumer; the stroma the producer. The phenomenon underpins much of the metabolic cross-feeding studied in the tumour microenvironment (Part VI).

9. Metabolic Heterogeneity and the Microenvironment

The single tumour is a metabolic ecosystem. In vivo infusion studies with [U-¹³C]-glucose and [U-¹³C]-glutamine in NSCLC patients ( Hensley et al. Cell 2016; Faubert et al. Cell 2017) overturned the reductive picture of “the” cancer metabolic phenotype. Within a single tumour, regions differ by:

• Oxygen tension — from near-zero in necrotic cores to perivascular normoxia.
• Glucose & glutamine availability — perfused vs hypovascular zones.
• Stromal composition — CAFs vs immune infiltrate vs adipocyte-rich margins.
• Genotype — subclonal driver mutations affecting metabolic enzymes themselves.

Cross-feeding examples

• Lactate shuttle (Section 8) — hypoxic exporter / oxic consumer.
• Adipocyte–tumour cross-talk in ovarian metastasis: omental adipocytes release fatty acids consumed by ovarian carcinoma cells via FATP/CD36.
• Asparagine cross-feeding from CAFs to asparagine-auxotrophic ALL/AML cells.
• Cystine–cysteine exchange via xCT (SLC7A11) protects tumour cells from oxidative death; ferroptosis vulnerability emerges when xCT is blocked.

Immunometabolic competition

Tumour cells are not the only consumers in the tumour. Activated effector T cells run their own glycolytic programme. Glucose depletion and lactate accumulation in the tumour interstitium directly suppress CD8⁺ T-cell function (Chang et al. Cell 2015). Lactate at >10 mM blocks Treg-suppressing FOXP3 dynamics and biases macrophages toward an immunosuppressive M2 phenotype. The metabolic microenvironment is therefore an immune environment as well — a theme taken up further in Part VI.

10. Targeting Cancer Metabolism

Two waves of metabolic-therapy development have come and gone, and a third is now in clinic. The first wave, in the 1940s–1960s, gave us folate and pyrimidine antimetabolites (methotrexate, 5-FU, gemcitabine) and l-asparaginase — tools still in front-line use. The second wave, in the 2000s, focused on the Warburg effect itself (DCA, 2-DG, 3-BP) and largely failed in trials due to narrow therapeutic windows. The current third wave targets specific oncometabolic vulnerabilities defined by genotype.

Mutant-IDH inhibitors — the success story

Ivosidenib (Agios, FDA 2018) and enasidenib (FDA 2017) demonstrated for the first time that targeting an oncometabolite-producing enzyme could induce terminal differentiation of leukaemic blasts — a remission mechanism reminiscent of ATRA in APL but achieved through 2HG depletion and TET2 reactivation rather than fusion-protein degradation. Vorasidenib’s 2024 approval for low-grade glioma (INDIGO trial: median PFS 27.7 vs 11.1 months) extended the strategy to a solid CNS tumour, exploiting the drug’s blood-brain-barrier penetration.

Glutaminase, FASN, LDH inhibition — ongoing battles

CB-839 (telaglenastat) is the clinical GLS1 inhibitor. In the ENTRATA trial it improved PFS in advanced clear-cell RCC when combined with cabozantinib, but the CANTATA trial failed its primary endpoint — illustrating how compensatory metabolic flux (glucose-derived TCA, pyruvate carboxylase up-regulation) limits monotherapy. TVB-2640 showed activity in HER2-positive breast cancer cells with high FASN dependence, and is in NASH Ph III. LDH-A inhibition produces convincing preclinical effects but no clinically validated compound has emerged — the active site is highly conserved with LDH-B and the substrate is small and polar.

Metformin — epidemiology meets metabolism

Diabetics on metformin have lower cancer incidence in some retrospective cohorts (~30% reduction in some analyses). Mechanistically metformin partially inhibits mitochondrial Complex I, raising AMP/ATP and activating AMPK, which inhibits mTORC1; it also suppresses gluconeogenesis and circulating insulin, indirectly limiting tumour growth factor signalling. Randomised oncology trials (METTEN, MA.32) have shown mixed results; the drug is cheap, safe and metabolically pleiotropic, which is both its attraction and its analytical headache.

Outlook

The next decade of cancer-metabolism therapy will likely depend on three advances: (1) genotype-matched assignment, as in the IDH paradigm, identifying tumours uniquely dependent on a metabolic enzyme; (2) combination strategies that close compensatory routes (GLS-i + glycolysis-i; FASN-i + SCD1-i); (3) the integration of metabolic targeting with immunotherapy, since lactate, kynurenine, adenosine and methionine in the tumour interstitium all condition the immune response. The conceptual journey from Warburg’s 1924 measurement to vorasidenib’s 2024 approval has taken exactly a century — and the territory between aerobic glycolysis and the epigenome is now firmly on the oncologist’s map.